Update app.py

app.py

@@ -2,124 +2,123 @@ import gradio as gr
 import pandas as pd
 from huggingface_hub import hf_hub_download
 
-# --- Configuration ---
+# --- Constants ---
 REPO_ID = "DontPlanToEnd/UGI-Leaderboard"
 FILENAME = "ugi-leaderboard-data.csv"
 
-# --- Constants for UI ---
-TITLE = "🏆 UGI Index Leaderboard"
-DESCRIPTION = """
-This is a modified version of the [UGI Leaderboard](https://huggingface.co/spaces/DontPlanToEnd/UGI-Leaderboard) that introduces the **UGI Index**.
-
-### 📊 The UGI Index
-The **UGI Index** is a custom score calculated to balance intelligence, uncensored knowledge, and willingness to answer.
-**Formula:** `(UGI Score + NatInt Score) * (W/10 Score)`
-
----
-**Original Metrics:**
-*   **UGI (Uncensored General Intelligence):** Measures knowledge of sensitive topics.
-*   **NatInt (Natural Intelligence):** Measures general reasoning and knowledge.
-*   **W/10 (Willingness):** Measures how far a model can be pushed before refusing.
-"""
-
 def make_clickable_model(model_name, link):
-    """Generates an HTML link for the model."""
-    if pd.isna(link) or link == "":
+    """
+    Wraps the model name in an HTML anchor tag if a link exists.
+    """
+    if pd.isna(link) or link == "" or str(link).lower() == "nan":
         return model_name
-    return f'<a target="_blank" href="{link}" style="color: var(--link-text-color); text-decoration: underline;text-decoration-style: dotted;">{model_name}</a>'
+    return f'<a target="_blank" href="{link}" style="color: var(--link-text-color); text-decoration: underline; text-decoration-style: dotted;">{model_name}</a>'
 
 def get_leaderboard_data():
     """
-    Fetches, cleans, and calculates the UGI Index.
+    Fetches the CSV, standardizes column names, calculates the UGI Index, and formats the table.
     """
     try:
-        # 1. Download CSV
+        # 1. Download the CSV
         file_path = hf_hub_download(repo_id=REPO_ID, filename=FILENAME, repo_type="space")
         df = pd.read_csv(file_path)
 
-        # 2. Clean Column Names
-        # The original CSV has trailing spaces (e.g., "UGI ", "W/10 ")
-        # We strip them for easier processing but keep a map if needed
+        # 2. Clean up column names (remove whitespace)
         df.columns = df.columns.str.strip()
 
-        # 3. Identify Key Columns
-        # We look for columns loosely to be robust against small upstream changes
+        # 3. Standardize the 'Model' column name
+        # Find whatever column looks like "Model" and rename it to "Model" to prevent KeyErrors
+        model_col_candidate = next((c for c in df.columns if "Model" in c), None)
+        if model_col_candidate:
+            df.rename(columns={model_col_candidate: "Model"}, inplace=True)
+        else:
+            # Fallback if no model column is found
+            df['Model'] = "Unknown"
+
+        # 4. Identify other specific columns dynamically
         ugi_col = next((c for c in df.columns if "UGI" in c and "Index" not in c), "UGI")
         natint_col = next((c for c in df.columns if "NatInt" in c), "NatInt")
         w10_col = next((c for c in df.columns if "W/10" in c), "W/10")
-        
-        # 4. Calculate UGI Index
-        # Formula: (UGI + NatInt) * (W/10)
-        # Convert to numeric, coercing errors to 0 or NaN to prevent crashes
-        c_ugi = pd.to_numeric(df[ugi_col], errors='coerce').fillna(0)
-        c_natint = pd.to_numeric(df[natint_col], errors='coerce').fillna(0)
-        c_w10 = pd.to_numeric(df[w10_col], errors='coerce').fillna(0)
+        link_col = next((c for c in df.columns if "Link" in c), None)
 
-        df['UGI Index'] = (c_ugi + c_natint) * c_w10
+        # 5. Ensure numeric data for calculations
+        for col in [ugi_col, natint_col, w10_col]:
+             # Force convert to numeric, turning errors (like strings) into NaN, then fill with 0
+             df[col] = pd.to_numeric(df[col], errors='coerce').fillna(0)
+
+        # 6. Calculate the UGI Index
+        # Formula: (UGI + NatInt) * (W/10)
+        df['UGI Index'] = (df[ugi_col] + df[natint_col]) * df[w10_col]
         df['UGI Index'] = df['UGI Index'].round(2)
 
-        # 5. Format Model Links
-        # If a 'Link' column exists, use it to make the 'Model' column clickable
-        if 'Link' in df.columns:
-            df['Model'] = df.apply(lambda row: make_clickable_model(row['Model'], row['Link']), axis=1)
-            # Drop the link column to keep the table clean
-            df = df.drop(columns=['Link'])
+        # 7. Make Model Names Clickable
+        if link_col and link_col in df.columns:
+            df['Model'] = df.apply(lambda x: make_clickable_model(x['Model'], x[link_col]), axis=1)
+            # Remove the Link column so it doesn't show up twice
+            df = df.drop(columns=[link_col])
 
-        # 6. Sorting
-        # Sort by UGI Index descending
+        # 8. Sort by UGI Index (Descending)
         df = df.sort_values(by='UGI Index', ascending=False)
 
-        # 7. Add Rank
+        # 9. Add Ranking
         df.insert(0, 'Rank', range(1, len(df) + 1))
 
-        # 8. Reorder Columns
-        # We want Rank, Model, UGI Index, then the core metrics, then the rest
+        # 10. Reorder Columns
+        # Ensure these main columns come first
         priority_cols = ['Rank', 'Model', 'UGI Index', ugi_col, natint_col, w10_col]
-        
-        # Get all other columns that exist in the dataframe
+        # Add whatever other columns exist in the CSV
         other_cols = [c for c in df.columns if c not in priority_cols]
         
-        # Combine
-        final_cols = priority_cols + other_cols
-        df = df[final_cols]
+        final_df = df[priority_cols + other_cols]
 
-        return df
+        return final_df
 
     except Exception as e:
-        return pd.DataFrame({"Error": [f"Failed to process data: {str(e)}"]})
+        return pd.DataFrame({"Error": [f"An error occurred: {str(e)}"]})
 
-def search_filter(query):
+def search_models(query):
+    """
+    Filters the dataframe based on the search query.
+    """
     df = get_leaderboard_data()
     if query:
-        # Case-insensitive search on the Model column
+        # We can now safely access 'Model' because we renamed it in the data loading step
         df = df[df['Model'].astype(str).str.contains(query, case=False, na=False)]
     return df
 
-# --- Application Layout ---
-with gr.Blocks(title="UGI Index Leaderboard") as demo:
-    gr.Markdown(f"# {TITLE}")
-    gr.Markdown(DESCRIPTION)
+# --- Gradio Interface ---
+custom_css = """
+footer {visibility: hidden}
+"""
+
+with gr.Blocks(css=custom_css, title="UGI Index Leaderboard") as demo:
+    gr.Markdown("# 🏆 UGI Index Leaderboard")
+    gr.Markdown("""
+    **A reordered view of the [UGI Leaderboard](https://huggingface.co/spaces/DontPlanToEnd/UGI-Leaderboard) based on the UGI Index.**
     
+    The **UGI Index** combines uncensored knowledge, general intelligence, and willingness to answer.
+    
+    $$ \\text{UGI Index} = (\\text{UGI Score} + \\text{NatInt Score}) \\times \\text{W/10 Score} $$
+    """)
+
     with gr.Row():
-        search_box = gr.Textbox(
-            label="Search Models", 
-            placeholder="Type a model name (e.g. Llama-3)...",
-            interactive=True
+        search_bar = gr.Textbox(
+            label="Search",
+            placeholder="Search for a model name...",
+            show_label=False
         )
-        refresh_button = gr.Button("🔄 Refresh Data")
+        refresh_btn = gr.Button("Refresh Leaderboard")
 
-    # The Dataframe component
-    # datatype="markdown" is crucial for rendering the HTML links
-    leaderboard_table = gr.Dataframe(
+    leaderboard = gr.Dataframe(
         value=get_leaderboard_data,
-        datatype="markdown",
+        datatype="markdown",  # Essential for rendering HTML links
         interactive=False,
         wrap=True
     )
 
-    # Wire up the interactions
-    search_box.change(fn=search_filter, inputs=search_box, outputs=leaderboard_table)
-    refresh_button.click(fn=get_leaderboard_data, outputs=leaderboard_table)
+    # Event Listeners
+    search_bar.change(fn=search_models, inputs=search_bar, outputs=leaderboard)
+    refresh_btn.click(fn=get_leaderboard_data, outputs=leaderboard)
 
-# Launch
-demo.launch()
+if __name__ == "__main__":
+    demo.launch()